Productive mRNA stem loop-mediated transcriptionalslippage: Crucial features in common withintrinsic terminatorsChristophe Pennoa, Virag Sharmaa, Arthur Coakleya, Mary O’Connell Motherwaya, Douwe van Sinderena,Lucyna Lubkowskab, Maria L. Kireevab, Mikhail Kashlevb, Pavel V. Baranova, and John F. Atkinsa,c,1

aSchools of Biochemistry & Microbiology, and Alimentary Pharmabiotic Centre, University College Cork, Cork, Ireland; bCenter for Cancer Research, NationalCancer Institute, Frederick, MD 21702; and cDepartment of Human Genetics, University of Utah, Salt Lake City, UT 84112

Edited by Jeffrey W. Roberts, Cornell University, Ithaca, NY, and approved March 3, 2015 (received for review September 24, 2014)

Escherichia coli and yeast DNA-dependent RNA polymerases areshown to mediate efficient nascent transcript stem loop forma-tion-dependent RNA-DNA hybrid realignment. The realignmentwas discovered on the heteropolymeric sequence T5C5 and yieldstranscripts lacking a C residue within a corresponding U5C4. Thesequence studied is derived from a Roseiflexus insertion sequence(IS) element where the resulting transcriptional slippage is re-quired for transposase synthesis. The stability of the RNA struc-ture, the proximity of the stem loop to the slippage site, the lengthand composition of the slippage site motif, and the identity of its3′ adjacent nucleotides (nt) are crucial for transcripts lacking asingle C. In many respects, the RNA structure requirements for thisslippage resemble those for hairpin-dependent transcription ter-mination. In a purified in vitro system, the slippage efficiencyranges from 5% to 75% depending on the concentration ratiosof the nucleotides specified by the slippage sequence and the 3′nt context. The only previous proposal of stem loop mediatedslippage, which was in Ebola virus expression, was based on in-correct data interpretation. We propose a mechanical slippagemodel involving the RNAP translocation state as the main motorin slippage directionality and efficiency. It is distinct from previouslydescribed models, including the one proposed for paramyxovirus,where following randommovement efficiency is mainly dependenton the stability of the new realigned hybrid. In broadening thescope for utilization of transcription slippage for gene expression,the stimulatory structure provides parallels with programmed ribo-somal frameshifting at the translation level.

transcriptional realignment | stem loop stimulator | heteropolymericslippage-prone motifs | frameshifting

Arelatively neglected aspect of gene expression is functionallyimportant RNA-DNA hybrid realignment within the RNA

polymerase (RNAP) transcribing coding sequence. Such real-igned polymerases yield transcripts lacking, or containing, one ormore additional base(s) corresponding to the slippage-pronesequence. The fraction of the mRNA population containing de-letions or insertions with respect to the template that are not of 3nucleotides (nt) or multiples thereof, yields proteins that are ei-ther truncated, have C-terminal regions not encoded by the zeroframe of the template DNA, or both. Despite the limited in-vestigations, several diverse functional slippage-derived productshave already been characterized.Bacterial multisubunit DNA-dependent RNAPs undergo

slippage on homopolymeric runs of As or Ts. Transcriptionalslippage on a run of nine As in the Thermus thermophilus dnaXgene results in heterogeneous transcripts that lead to 50% of theribosomes quickly terminating. This termination yields a DNApolymerase subunit that lacks C-terminal domains and is syn-thesized in a 1:1 ratio with the full-length product (1). In otherbacteria, counterpart slippage is required for synthesis of the full-length rather than the truncated product (2). Disease significanthomopolymeric transcriptional slippage occurs in expression of

the secretion system of Shigella flexneri (3, 4), Citrobacter, andYersinia (5). Mechanistically similar slippage is involved inseveral bacterial IS element transposases (6) and probably inStaphyloccocus aureus mapW (7). Striking instances of transcrip-tional realignment occur in gene expression of endosymbioticbacteria of insects that have undergone rapid evolution resultingin high AT content and exceptional genome reduction (8, 9).The length of the hybrid between nascent RNA and its tem-

plate within a transcribing RNAP is highly relevant to slippagemotif length. The length of the nascent RNA template hybridoscillates by one base during nt cycling in the process of baseaddition. There is a longer hybrid when the RNAP is in thepretranslocated state with the catalytic site located at the RNA 3′end (termed position i in its DNA template). The shorter hybridis when the RNAP is in the posttranslocated state with the cat-alytic site vacated from the 3′ RNA end and ready for binding ofthe next cognate nucleoside triphosphate (NTP) (specified by DNAtemplate position i + 1) (10). For Thermus thermophilus in whichdnaX slippage was discovered, crystallographic structures haverevealed that the respective lengths are 9 and 10 bp (11). Cor-respondingly, slippage occurs efficiently (can be ∼50% depend-ing on the organism and 5′ nt context) with 9 As or Ts but isgreatly reduced with shorter runs (12–14).Functionally important slippage of viral-encoded RNA-

dependent RNA polymerases has been studied in the expression

Significance

Perturbation of transcription register by RNA polymerase, tran-scription slippage, is used to yield additional protein products.Known functionally important cases involve a small number ofshort sequences without secondary structure. The discoveryreported here of the dependence of a newly identified motif onnascent RNA forming a stem loop structure within the RNA exitchannel of the polymerase greatly extends the potential for abroad variety of putative slippage sequences, especially as thephenomenon has been observed with both bacterial andeukaryotic RNA polymerases. Characterization of the mechanisminvolved shows similarities with, and differences from, intrinsictranscription termination, which also depends on formation ofRNA stem loop structures. Our findings reveal novel insights toRNA polymerase versatility and functioning.

Author contributions: C.P., L.L., M.L.K., M.K., and J.F.A. designed research; C.P., A.C.,M.O.M., and L.L. performed research; C.P., L.L., M.L.K., and M.K. performed in vitro ex-periments; C.P., V.S., M.O.M., and D.v.S. contributed new reagents/analytic tools; C.P.,A.C., D.v.S., L.L., M.L.K., M.K., P.V.B., and J.F.A. analyzed data; C.P., V.S., M.L.K., M.K.,and J.F.A. wrote the paper; and J.F.A. provided overall direction.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: j.atkins@ucc.ie.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1418384112/-/DCSupplemental.

E1984–E1993 | PNAS | Published online April 6, 2015 www.pnas.org/cgi/doi/10.1073/pnas.1418384112

Dow

nloa

ded

by g

uest

on

May

24,

202

1

of the Filovirus, Ebola, where slippage occurs on a run of 7As,and in the expression of Paramyxoviruses, especially Sendai virus(15–17). Note that in negative strand virus research, synonymsfor transcription slippage, such as transcriptional stuttering,cotranscriptional editing, pseudotemplated transcription, or re-iterative transcription, are frequently used (18). Unlike bacterialRNAPs in the examples described above, some viral RNAPsstudied slip on the heteropolymeric sequence AmGn (where mand n are integers). On this motif the polymerase from severalParamyxoviruses add a nontemplated G nucleotide(s). The ef-ficiency of G addition is 31% for Sendai and 82% for Nipah (19).The number of G insertions differs in different Paramyxovirussubfamilies (one in respiratory/morbilli and two in the rubulasubfamily) and is dependent on the sequence upstream of theslippage site (20). The number of added Gs is correlated to theorganization of the P gene in which it occurs in the different viruses(see ref. 21 for a discussion of the linkages of viral features).As the propensity for slippage is strongly influenced by the

stabilities of the initial and newly realigned nascent RNA-tem-plate hybrid (22, 23), slippage on heteropolymeric sequences isespecially sensitive to realigned base pairing strength. Use of theAmGn motif involves formation of nondestabilizing rG:rU basepairs, but not rA:rC, and thus confers realignment directionality.With a motif containing shorter G tracts, it is only G addition(s)that occur and its efficiency is inversely proportional to the sta-bility of the newly realigned RNA-RNA hybrid. In addition, in-creasing the length of the A tract of the motif leads to A addition(24). In contrast, with a motif containing a longer G tract, Gdeletion(s), but not addition, occurs (25).A previous bioinformatic analysis of disrupted bacterial ORFs

revealed that bacterial RNAPs might also slip on heteropolymericmotifs. Conserved XmYn motifs were identified as candidatesfor productive transcriptional realignment required for synthesisof a product encoded by two partially overlapping ORFs (26).One potential slippage site is in the 80 nt overlap region of thefirst (272 nt) and second (966 nt) ORFs of the IS630 family in-sertion sequence in Roseiflexus sp, RS-1 (Fig. 1). The secondORF is in the +1 translational reading frame with respect to theframe in which translation initiates. The trans-frame encodedprotein is the functional transposase. Prior analysis of counter-part instances of where synthesis of a trans-frame encodedtransposase requires a ribosomal frameshifting event has shownthat the frameshifting is necessary for transposition (27, 28).Here we investigate whether the product encoded trans-framewith respect to the DNA sequence derives from transcriptionalslippage or ribosomal frameshifting on the heteropolymeric T5C5motif and the characteristics of the mechanism involved.

ResultsAn Inverted Sequence Is Required for Transcriptional Slippage on aT5C5 Motif. To investigate whether Escherichia coli RNAP un-dergoes slippage within the T5C5 motif in the Roseiflexus IS630ORF1 ORF2 overlapping region (Fig. 1), we used the gst-mbpexpression plasmid system pJ307 (Fig. 2 A–C). A 41 nt cassettefrom the overlap region that contains 30 nt 5′ of the T5C5 motifand its 3′ flanking nt, was fused between the reporter encodingsequences gst (in-frame) and mbp (−1 frame). In the non-template strand of IS630, the T5C5 motif is 5′ adjacent to theORF1 stop codon (Fig. 2D, insert 1). Analysis of proteins expressedin E. coli shows 20% efficient ORF switching (Fig. 2E, insert 1)from the 0 [encoding the standard glutathione S-transferase(GST) product] to the +1 reading frame [yielding the trans-frameGST-maltose binding protein (MBP) product]. To determine iftranscription slippage event(s) are occurring, the correspondingtranscription products were analyzed by limited primer extension(LPE). This technique involved annealing labeled primer to se-quence 3′ adjacent to the slippage motif, extending the primerusing a set of three dNTPs with the fourth and missing onereplaced by its corresponding chain terminator dNTP derivative.Only one primer extended product is expected to be delimited 5′by the 5′ end of the primer and 3′ by the first template locationspecifying incorporation of the chain terminator (Fig. 2F; L acyAand L acyC show two independent reactions using acyclonucleotideacyA or acyC as the chain terminator). Specific addition, or lack ofnt(s) corresponding to a template nt, before incorporation of thechain terminator nt, results in a longer or shorter LPE product(s).In this study, transcriptional slippage analysis involved comparisonof the LPE products generated from the RT-PCR template fromthe mRNA population, designated T (Test) and generated fromPCR template DNA, designated L (Ladder), to reveal any DNApolymerase slippage during PCR that could mislead interpretationof the results (Fig. 2, in green). LPE on the RT-PCR productderived from cells expressing a reporter carrying the original insert(Fig. 2F, insert 1 T) results in two products, the longer of whichcorresponds to the RT-PCR carrying T5C5 (as specified by thetemplate) and the shorter one corresponds to the RT-PCR carry-ing T5C4. The corresponding control (Fig. 2F, insert 1 L) showedonly one product, which is the same size as the longer one presentin T, i.e., indicating homogeneity T5C5 in the PCR product.Therefore, LPE analysis established that 30% of the RT-PCRproduct derived from the WT IS630 41nt cassette lacked one cy-tosine in the C5 tract (Fig. 2F, insert 1).Next, we analyzed if the sequence upstream of the T5C5 motif

is required for robust transcription slippage, leading to the Cdeletion. In WT mRNA (insert 1), there is a 6 nt inverted repeatsequence with potential to form a stem loop structure containinga GCAA loop 5′ adjacent to the U5C5/U5C4 motif (Figs. 1 and2D, insert 1). Deletion of the 12 nt 5′ of the inverted repeatsequence caused only a marginal reduction in slippage efficiency(Fig. 2 D and F, insert 2), but deletion of 20 nt that included the6 nt in the 5′ side of the stem abolished slippage (Fig. 2 D and F,insert 3). In addition, substituting the sequence in the 5′ side ofthe stem with its complement, gcgggc to cgcccg, removed basepairing potential and also abolished slippage (Fig. 2 D and F,insert 4).To test the possibility of reverse transcriptase slippage that

would generate heterogeneity in the RT-PCR product, LPE re-actions were also performed on chemically synthesized RNA,designated S (Synthetic; Fig. 2, in green). For that, we used theminimal sequence of the Roseiflexus slippage-prone cassette thatexhibits strong slippage as detected by LPE analysis (Fig. 2D,insert 2 template specifies the inverted sequence and the U5C5motif region). We conclude that reverse transcriptase slippagedoes not slip on the U5C5 motif and LPE faithfully detectsthe lack of one C in the mRNA compared with its template(Fig. 2F, Lower).Transcription slippage resulting in the lack of one C in the

transcript is expected to yield a full-length GST-MBP fusion pro-tein (Fig. 2C). The GST-MBP protein expression was correlated

A

B

C

Fig. 1. Roseiflexus IS630 insertion sequence. (A) Diagram of the IS630 familyinsertion sequence transposase gene that contains two ORFs. The 3′ end oforf1 overlaps the 5′ end of orf2 which is in the +1 frame with respect to it.The nucleotide sequence displayed is of the overlapping region of the twoorfs (coordinates 561509–561588 in RefSeq entry NC_009523.1). The invertedsequence 5′ of the underlined T5C5 motif is indicated by arrows. (B) orfconfiguration in transcripts generated without or with slippage. (C) Productsfrom standard translation of these transcripts.

Penno et al. PNAS | Published online April 6, 2015 | E1985

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

24,

202

1

with the ability of nascent RNA to potentially form a stem loopstructure. The GST-MBP fusion product comprised 20% and 15%of the total reporter protein (GST and GST-MBP) expressed fromthe constructs containing inserts 1 and 2, respectively (Fig. 2 D andE); the fusion product was not detected for inserts 3 and 4 (Fig. 2D and E). A similar result was also obtained in vivo with the gstlacZ encoded reporters for inserts containing the WT (inserts1 and 2) or mutant (inserts 3 and 4) sequence (SI Results and Fig.S1). Taken together, the analyses of protein product(s) and tran-scripts synthesized in vivo show that trans-frame encoded proteinsynthesis is at least mainly driven by transcription slippage, whichled to the programmed absence of one C in the mRNA. In con-clusion, the minimal T5C5 slippage-prone sequence is delimited 5′by the inverted repeat sequence and 3′ by the stop codon (insert 2).The sequence of insert 2 was used as the WT slippage-inducingcassette in the following experiments.

Features of the Stem. The strength of base pairing potentialspecified by the inverted repeat sequence (gcgggcGCAAgcccgc,with sequence of the stem indicated in lowercase) was in-vestigated with constructs derived from insert #2 (Fig. 2D). First,a base pairing variant indicated that slippage occurs when GC-rich base pairing is maintained (insert 5, cgcccgGCAAcgggcg,

switching the 5′ and 3′ sides of the stem) but does not occur withG:U (insert 6, guggguGCAAguuugu, C replaced by U) or A:Ualternatives (insert 7, auaaauGCAAauuuau, C to U and G to A;Fig. 2 D and F, inserts 5–7), supporting the idea that the originalinverted repeat sequence forms a stem loop structure.Next, a G-to-A substitution at the first position of the 3′ side of

the stem that disrupts potential base pairing gave 25% slippage(Fig. 3, insert 45). Changing both the first and second base on the3′ side of the stem to A further disrupts potential pairing to givethe same 8 base loop as in insert 41 in the loop analysis below. Itresulted in just 5% slippage (insert 46). Changing the 5′ base Gon the 5′ side of the stem to C gave 10% slippage (insert 47).Replacing both the 5′ G and C at the base of the 5′ stem alsoresulted in 10% slippage (insert 48). Substituting the C 5′ adja-cent to the stem with A creates the potential for a 1-bp stembottom extension by pairing with the first U of the slippery motif.This construct (insert 52) has the same slippage efficiency as WT.Creating the potential for an extra CG base pair at the top of thestem led to slippage being reduced from 30% to 25% (insert 53).A combination of these last two changes creating the potentialfor two extra base pairs did not change the efficiency from WT(insert 54). Creating the potential for a second extra base pair atthe top of the stem did not change the efficiency (insert 55).

A

B

C

E

D

F

Standard

Fig. 2. Stimulatory effect of Roseiflexus stem loop on T5C5 motif slippage. (A) pJ307 vector with fused gst mpb reporter genes separated by the test insertwhose propensity for mediating reading frame changes is monitored for standard (U5C5) transcription or that involving a realignment event (U5C4) (B), leadingto the synthesis of GST or transframe encoded GST-MBP products (C). Right pointing arrows from A and B show the steps used to generate the limited primerextension products, L, ladder, to monitor potential replication slippage; T, Test, to monitor transcriptional slippage; and S, synthetic, as control for reversetranscriptase slippage. (D) (Upper) Sequence of insert 1 and its derivative inserts 2–4. (Lower) Predicted RNA stem loop structure of the shorter WT sequence ininsert 2 and its variants 5–7. (E) Pulse chase analysis of proteins decoded from pJ307 (A) with test cassettes. GST reports zero frame translation and, with testsequences, GST-MBP reports products that are transframe encoded with respect to the DNA sequence; 16 and 15 represent in-frame and out-of-frame controls,respectively. SEM error for frameshifting efficiency was <25% for three independent experiments. (F, Upper) Expected LPE products. The primer is indicated bya green arrow. As illustrated by the length of the different reaction products terminated for the Ladder “L” after the C5 tract (acyA terminator) or T5C5 motif(acyC terminator), changes at any nt position before the termination site can be identified. (Lower) DNA sequencing gel analysis of LPE reaction products. L(Ladder) and T (Test) signify limited primer extension based on template sequence using PCR or RT-PCR, respectively. S (Synthetic) is as T but using chemicallysynthetized RNA. Acyclonucleotides, acyA and acyC, that cause termination, are indicated on the left; they reveal base insertion or base absence for the se-quence containing the 3′ part of motif (5′-C5-3′) and the whole motif (5′-T5C5-3′). The termination sites of the primer extension, corresponding to the cDNAtemplate base of the acyclonucleotide, are shown for inserts 1–7 that are diagrammed in Fig. 3. In the PCR-derived ladder (L), only one product indicatesabsence of slippage at any level.

E1986 | www.pnas.org/cgi/doi/10.1073/pnas.1418384112 Penno et al.

Dow

nloa

ded

by g

uest

on

May

24,

202

1

A B C

D E F G H I

Fig. 3. Constructs, slippage analysis, and RNA stem loop-related structures. (A) Diagram of important features for TmCn derived Roseiflexus slippage-pronecassettes, displayed above the mutant classes with their insert numbers. (B) Insert sequences. The reading frame of the 5′ reporter gene, gst, is indicated at theleft with the underlined TCG, which is part of the restriction site used for cloning. The in-frame stop codon, TAG, which is part of the 3′ restriction cloning site,is in bold. It overlaps with the underlined AGA that indicates the frame of the second reporter. Sequences potentially able to form a stem are in green withtheir loop sequence in italics. The Ts of TmCn potential slippery motifs are in blue and the Cs in red. Substitution mutant sequence(s) are in lowercase andbold; addition(s) are in purple; hyphens (-) indicate nt absence. The purple arrow at the bottom of the stop codon containing sequence column indicates the 3′annealing site for limited primer extensions, with a specific dNTP replaced by the corresponding terminator acyNTP. This reveals the termination site nts whichare shown boxed. Red box, position 1, shows the termination site used to identify transcripts whose sequence for the second half of the TmCn motif, i.e., Cn,has 1 base more or less than in the template. Blue box, position 2, shows the counterpart termination site for the whole motif. Absence of a box means theLPE analysis was not determined. (C) Relative yield of products with more or less nucleotide(s) than the template. Each length unit, as indicated by smallvertical bars, corresponds to 5% slippage efficiency. When slippage was involved, the RT-PCR products were analyzed on at least on three independentsamples of in vivo RNA (and on two when it was not). Error bars: SEM. (D–G) alternative RNA stem loop(s), inserts 49 and 61–62. (H) Known histidine pauseattenuator 5′ of the RNA 3′end pause site. (I) T3 intrinsic terminator of the metY-nusA-infB operon (49).

Penno et al. PNAS | Published online April 6, 2015 | E1987

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

24,

202

1

Adding to this the potential extra base pair at the bottom (as ininsert 52) gave 27% slippage, similar to WT (insert 56).Substituting the first 3 nt on the 5′ side to the stem to their

complementary sequence CGC caused 15% slippage, perhapsdue in part to alternative pairing (insert 49; Fig. 3D). Disruptionof the 3 bp at the base of the stem by mutations at the 3′ side ofthe stem abolished slippage (insert 50). Interestingly, combiningreplacements from insert 49 and insert 50, to restore base pairingwithout affecting G/C content of the stem (5′CGC/5′GCG), gives10% slippage (insert 51), which is three times less than the WT(5′GCG/5′CGC) 30% level. This result indicates that, in additionto the base pairing in the stem, the 3′ CGC sequence in RNA orDNA has a separate contribution to slippage (Discussion). Ac-cordingly, swapping the left and right arms of the stem (insert 5),without changing the loop, decreased slippage from 30% to 20%(compare inserts 2 and 5).

Features of the Loop. The loop, 5′-GCAA-3′, is a GNRA-typetetraloop (with the nature of the four bases being: G, guanine; N,any; R, purine; A, adenine) (29), provoking an analysis of thepotential importance of the nature of the loop. Substituting thesebases by their complimentary bases (5′-CGTT-3′) decreasedslippage slightly to 25% (Fig. 3, insert 39). Changing the 3′ baseto C also led to a similarly moderate decrease (insert 40). A 4-bpinsert to create 5′-GCGCAAAA-3′ also led to 25% slippage (insert41). Changing to the different GNRA sequence, 5′-GAAA-3′,yielded a modest increase to 35% (insert 42). Substitution withthe most common tetraloop, the sequence in Rho-independentstem loop transcription terminators, 5′-TTCG-3′, also led to amodest increase to 35% (insert 43). Replacement with the loopof the histidine operon attenuator, pause stem loop, 5′-CTAA-GTCTT-3′, led to a slight decrease to 25% (insert 44).The corresponding LPE gel analysis is shown in Fig. S2 A

(stem mutants) and B (loop mutants). In summary, stability ofthe stem is important, although it can have latitude in its length,and more stabilizing loops had a modestly enhancing effect onslippage compared with the other loop sequences tested.

All 10 nts in the T5C5 Motif Are Required for Slippage. Structuralstudies of Thermus thermophilus RNAP provide strong evidencefor the length of the RNA-DNA hybrid during elongation os-cillating between 9 and 10 nt during forward and backwardtranslocation of the enzyme on DNA (11). One hypothesis toexplain the absence of 1 cytosine (5′-U5C4-3′) in a sizeable frac-tion of the mRNA (5′-U5C5-3′) transcripts is that a realignment ofthe nascent 3′ RNA end from its template DNA hybrid occurs by a1 nt forward shift.To ascertain the relative importance of the different motif

positions, T1T2T3T4T5C1C2C3C4C5, substitutions were made that,where possible, did not change the identity of the encoded aminoacids (Fig. 3, inserts 8–20). T1 was changed to A, C, or G, withresulting reduction to no more than 8% slippage (inserts 8–10).No slippage was detected with T4 changed to C (insert 11), T5 andC1, were together changed to A and G (insert 17), C2 waschanged to A, G, or T (inserts 12–14), and C5 was changed to A,G, or T (inserts 18–20; Fig. S2C, Left). The proteins synthesizedfrom these constructs were analyzed by a pulse chase experimentthat indicated no significant synthesis of a trans-frame slippageproduct except when the mutated motif contains at least three Cs5′ adjacent to the stop codon. The latter creates a “shifty stop”(30) for +1 ribosomal frameshifting (Fig. S2D).Successive deletion of the 5′Ts (inserts 21–23) and successive

deletion of the 3′Cs (inserts 24–26) abolished slippage (Fig. 3and Fig. S2C, Right). In summary, the slippage-prone sequenceinvolves all 10 nt of the motif and all motif mutations competentfor slippage strictly yielded transcripts with only a single C de-letion and no insertions or multiple deletions.

Slippage on T5C5-Derivative Motifs. The lack of one C in theslippage product derives from forward realignment of the RNAtranscript with respect to the DNA template resulting in an

rU:dG mismatch within the RNA-DNA hybrid. The potentialinfluence of the position of the rU:dG mismatch within TmCnmotifs was analyzed by successive replacements of T by C and ofC by T at the junctions in TmCn-containing constructs wherem + n was maintained at 10. From T1C9 to T7C3, the WTslippage efficiency was maintained (Figs. 3 and 4A, inserts 27, 28,and 31–34). From T8C2 to T9C1, the number of cytosines wasthe same as in the template DNA (inserts 35 and 38). Fivepercent of the transcripts from T9C1 contained an additionaluridine. This addition indicates a switch of the TmCn slippage toa different mechanism unrelated to programmed single C deletionwith respect to the template sequence. Pulse chase analyses ofprotein products showed a corresponding +1 trans-frame–encodedprotein whose relative abundance correlated with the proportionof transcripts lacking one C (U5C4; Fig. 4B, inserts 27, 28, 31–35,and 38). For T2C8 and T8C2, protein markers were provided byin-frame (inserts 30 and 37) and out-of-frame controls (inserts 29and 36; see insert sequences in Fig. 3). In conclusion, rU:dGmismatches located at positions from −3 to −9 in the new RNA-DNA hybrid do not prevent slippage (Fig. 4A, cartoon on bottom).The effect of increasing the number of Ts or Cs in the T5C5

motif was also analyzed (Fig. 3 and Fig. S2E, inserts 63–68). AtT5C6 and T5C7 sequences, slippage is at the WT level of ∼30%(inserts 66 and 67). A control for T5C7 with precluded stemformation gave 15% slippage (insert 68), showing that this motifon its own was now slippery. However, with T5C7 (insert 68), wecould not completely exclude the possibility of an alternativesecondary structure in the RNA that replaced the original stemloop. T6C5 and T7C5 also yielded WT slippage levels (inserts 63and 64). The control for T7C5 resulted in 15% slippage (insert65), which is the same as the T5C7 control (insert 68). In con-clusion, increasing the length of the TmCn slippery sequence(with m + n > 10) leads to slippage yielding a single C deletionwhose efficiency is dependent on the upstream RNA stem loop.

The Identity of the Base 3′ Adjacent to the T5C5 Motif Is Essential forSlippage. The nature of the first template base immediatelydownstream from the RNA-DNA hybrid affects RNAP pausing,forward translocation, and fidelity (31, 32). Its possible signifi-cance in the slippage-mediated origin of transcripts lacking asingle C was investigated using four constructs. These constructs

A B

Fig. 4. Stimulatory effect of Roseiflexus stem loop on TmCn motif slippage.(A) LPE analysis with similar designations to Fig. 2F. L (Ladder) and T (Test)refer to LPE analysis on PCR or RT-PCR template, respectively. The cartoon atthe bottom for the 9nt RNA-DNA hybrid shows the T/C junctions corre-sponding to each TmCn motif on the top of the gel. (B) Protein pulse chasecharacterization. The four lanes, inserts 29–30 and 36–37, represent out-of-frame and in-frame controls, for inserts 28 (T2C8 motif) and 35 (T8C2 motif)respectively. The insert sequences and corresponding slippage efficienciesare in Fig. 3. SEM error for frameshifting efficiency was <25% for three in-dependent experiments.

E1988 | www.pnas.org/cgi/doi/10.1073/pnas.1418384112 Penno et al.

Dow

nloa

ded

by g

uest

on

May

24,

202

1

were derived from the reference WT construct (insert 2) in whichthe T5C5 motif is 5′ to the T residue as part of a TAG stop codon.An additional base, N, was inserted between the T5C5 and theTAG (Fig. 3, inserts 66 and 69–71). Construct T5C6 (insert 66)had N as C and, as mentioned above, had the same slippage ef-ficiency as T5C5. With N instead being A, G, or T (inserts 69–71),LPE analysis showed that slippage was reduced from 30% with Cto 3%, 2%, and 10%, respectively (Fig. 3 and Fig. S3). Comparedwith the original T5C5 construct with a T 3′ adjacent to the motif(30% slippage), the above constructs contain a one base insert (N)3′ of the T5C5 motif with TAGATCC becoming NTAGATC.Therefore, an influence of the second 3′ (underlined) and perhapsthe following base(s) on slippage is not precluded (SI Discussion).In conclusion, the identity of the base 3′ adjacent to the T5C5motif influences slippage, with a purine having the strongestinhibitory effect.

Positioning of the Stem Loop Relative to the T5C5 Slippery Site. ForRNA structure-mediated effects on transcription terminationand on pausing, the positioning of the structure strongly in-fluences its effectiveness (33, 34). To assess for possible similarspacing influences for the Roseiflexus stem loop effect on slip-page, the distance between the stem loop and the slippage sitewas increased (Fig. 3, inserts 57–62). Adding one (insert 57),three, or four As eliminated slippage (inserts 59–60), althoughtwo As gave 2–3% slippage (insert 58). With the addition of onlyGCCCGC, the same sequence as the 3′ side of the stem, slippagewas 22% (insert 61), which was moderately reduced from WT(30%); in this case, an alternative stem loop with two extra basepairs (at its top) could form (Fig. 3E). With an additional repeatof this sequence, i.e., GCCCGCGCCCGC, the slippage leveldropped to 2–3% (insert 62), which is the same as merely addingtwo As (insert 58). In this case, two alternative stem loops mightform: one with three extra base pairs (two at its top and one at itsbottom; Fig. 3F) and the other with two extra base pairs (at itstop; Fig. 3G). The corresponding LPE gel analysis is shown inFig. S2E. In conclusion, the distance between the RNA stemloop and the slippery motif is crucial for programmed slippageleading to one C deletion.

Mechanistic Analyses of Slippage on T5C5. We used a minimal invitro transcription system involving ternary elongation complexes(TECs) reconstituted from RNAP core enzyme and syntheticRNA/DNA oligonucleotides (35). This system was used to testwhether slippage on the T5C5 motif depends only on the in-teraction of RNAP with the transcript and template in the 9- to10-bp RNA-DNA hybrid as has been shown for slippage onhomopolymeric tracts (22, 36). It was also used to confirmstimulation of slippage by the 5′ stem-loop structure. RNApolymerase was positioned upstream of the T5C5 slippery sitewith the 22 nt internally labeled nascent transcript with WT or amutated 5′ sequence of the stem (mut). Transcription was re-sumed by addition of unlabeled UTP, CTP, and GTP for theT5C5tga construct or ATP, CTP, and UTP for the T5C5tagconstruct. Lack of ATP or GTP in the reactions induced RNAPstalling 2 bp downstream from the end of the slippery tract (thebase highlighted in bold; Fig. 5A). The concentration of CTP wasvaried in the 5- to 500-μM range to investigate the effect onslippage of slow or fast transcription within the C5 part ofthe motif.Slippage on the T5C5 motif leading to lack of one C in the

transcript occurred in vitro during transcription of a syntheticDNA template in the absence of any additional transcriptionfactors. Changing the sequence 3′ to the slippery site, TAG, toTGA did not affect slippage (Fig. 5B), suggesting that G or Alocated 2 and 3 nt downstream from the T5C5 motif played norole in slippage. The efficiency of slippage on the template withthe intact stem loop inversely correlates with the concentrationof CTP in the reaction mix: lowering the CTP concentration ledto a larger fraction of the transcript lacking one C (Fig. 5B, WT).Consistent with the in vivo findings, slippage on the template

specifying RNA without the stem loop was observed at a very lowlevel and only at the lowest CTP concentration tested (Fig. 5B,mut), confirming the strong stimulator role of the stem loopstructure in slippage established in vivo.Changing the UTP/CTP ratio showed a dramatic effect on

slippage efficiency (Fig. S4 A and D). Although the UTP con-centration also affects transcription of the T5 part of the T5C5motif, it seems unlikely that this effect alters slippage in the C5part of the motif. Such realignment must occur with the initialRNA 3′end having no more than four C residues, i.e., located 4base pairs downstream from the T5 tract. Therefore, the UTP/CTP ratio effect indicates that a high rate of incorporation ofUTP, the nucleotide specified immediately downstream of themotif, is crucial for the lack of one C by slippage. Thus, thedependence of slippage efficiency on transcription rate appearsquite complex: slow transcription through the C5 part of theT5C5 motif promoted slippage, whereas slow addition of theNMP 3′ adjacent to the T5C5 motif inhibited slippage.Transcription termination was observed at positions C2 and

C3 (corresponding to the RNA 3′end sequence U5C2 and U5C3)with the test (WT) specifying the WT Roseiflexus structure, but notthe control (mut), which specifies a mutated structure (Fig. S4 Aand B). To test whether RNA stem loop formation has a similareffect on slippage as it does in transcription termination in desta-bilizing the closest part of the hybrid, phage lambda N protein,which stabilizes the upstream part of the hybrid (23, 37), was in-cluded in the assay.The presence of N protein caused the efficiency of stem loop-

dependent slippage on T5C5 to be reduced by ∼25–50% depending

0

10

20

30

40

50

60

70

80

90E. coli

S. cerevisiae

T5C5tag T5C5tgaslippery sitestructure wt wt wtmut mutmutN protein + + ----- -

C 5μMC 50μMC 500μM

% s

lippa

ge to

lack

one

C

U5C4UgU5C5Ug

T5C5tgacT5C5tg

Initial TEC

+ NTP

Final TEC

g

ca

cRNAtemplate

non-template 5’ 5’5’5’3’

3’ 3’3’

5’end5’end

3’ end 3’ end

A

B

C

Fig. 5. In vitro transcriptional slippage in a minimal purified system. (A) Ex-ample of in vitro slippage assay with the T5C5tga sequence. A TEC, con-taining RNAP, DNA (template and nontemplate strand), and nascent RNA,was generated at DNA template position 3′ adjacent to the 3′A5G5A5′ DNAtemplate (Initial TEC). Then, transcription elongation progressed to the DNAtemplate position 5′adjacent to template 3′A5G5A5′ sequence using a setof mixes of three NTPs including CTP at different concentration (Final TEC).(B) Slippage efficiency with RNAP purified from E. coli or RNAP II fromS. cerevisiae. The DNA template used yields the WT or mutated (mut)Roseiflexus RNA structure. The T5C5 motif is 5′ adjacent to TAG or TGA. For E.coli polymerase, the effect of N protein from phage lambda added at the InitialTEC stage is indicated. SEM error bars are indicated for three independentexperiments. (C) Representative sequencing gel of E. coli RNAP-derived tran-scripts from T5C5tga-containing templates. Part of the RNA sequence ofstandard (5′U5C53′) and slippage (5′U5C43′) products is shown on the Left.

Penno et al. PNAS | Published online April 6, 2015 | E1989

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

24,

202

1

on CTP concentration (Fig. 5B). This reduction is consistentwith the Roseiflexus stem loop directly mediating slippage viamelting of the closest part of the RNA-DNA hybrid or with Nprotein inhibiting hybrid realignment involving “defect migra-tion” (SI Discussion). A possible, perhaps additional, indirecteffect via stem loop formation associated pausing stimulatingslippage by enhancing the time window for hybrid realignmentwas explored. Pausing can be due to any of a number of reasons.In vitro experiments (above) showed the effect of substratelimitation. A further in vitro experiment used NusA transcriptionfactor that stimulates RNAP pausing at the histidine biosyntheticattenuator (38). It increased RNAP pausing at position U 3′adjacent to the motif but showed no effect on slippage (Fig. S4 Aand C). These finding are in agreement with in vivo data (SIResults) using E. coli rho and rho nusA strains, the propensity ofRNAP mutants, rpoB8 (slow elongation rate) and rpoB3595 (fastelongation rate), and the introduction of the “histidine pause”stem loop upstream of the T5C5 motif (Fig. 3H and inserts 72–73). Taken together, the in vitro and in vivo data do not reveal anindirect pausing-related mechanism of stem loop involvement inslippage stimulation.

Eukaryotic RNAP-Mediated Slippage. Promoter-independent invitro transcription permitted testing of whether eukaryotic RNApolymerase II (RNAP II) (35) also mediates stem loop-dependentslippage on T5C5. RNAP II from Saccharomyces cerevisiae alsoyielded transcripts lacking one C, although the proportion wasslightly less than with E. coli RNAP under the same conditions (Fig.5B, compare E. coli and yeast RNAPs for T5C5tga motif with WTRNA structure). Importantly, slippage efficiency of RNAP II wassignificantly stimulated by the RNA stem loop structure (Fig. 5B),and slippage was also sensitive to cytosine triphosphate (CTP)concentrations. To summarize, the T5C5 motif represents a uni-versal slippery sequence for distantly related RNAPs, and bothenzymes similarly use the stem-loop structure for slippage.

DiscussionThis work shows that E. coli RNAP undergoes transcriptionalslippage leading to lack of one cytosine in the RNA transcribedfrom a heteropolymeric T5C5 motif and certain derivatives.Slippage on heteropolymeric motifs during transcription elon-gation has been previously reported only for paramyxovirusRNA polymerases and involves AmGn motifs leading to addi-tions of G (21). Formation of an RNA stem-loop structurestrongly stimulates transcript realignment on a T5C5 motif. Theonly prior observation of transcriptional slippage stimulated by astem loop structure was reported for Ebola virus RNA-dependentRNA polymerase (39). We assert that no credible evidence hasbeen presented that it is a real precedent (SI Discussion). Impor-tantly, the RNA structure-dependent transcript realignment on theT5C5 tract is not limited to bacterial transcription. Yeast RNAP II,whose protein contacts with the nascent RNA and the RNA-DNAhybrid are similar to, but not identical with, its bacterial counterpart,efficiently slips in vitro in a hairpin-dependent manner on the T5C5motif. Potential relevance of polymerase structural differences todownstream DNA context on slippage efficiency merit future ex-ploration. It could be significant because with E. coli RNAP slip-page in vivo, the WT construct (insert 2) exhibited 30% slippage,whereas construct 71, which has the same 3′nt but contains an extra3′ flanking T, showed 10% slippage (Fig. 3 and SI Discussion).Several key features of the TmCn motif-containing cassette for

E. coli mRNA transcripts lacking a single C are latitude of theTmCn motif involving 2 < m<7 and n > 2 with a minimal lengthof m + n = 10, involvement of a Tm/Cn (5′T/C3′) pyrimidine/pyrimidine junction within the motif, presence of a pyrimidineresidue (Py, C, or T) 3′ adjacent to the TmCn motif, formation ofa G/C-rich stem loop structure 5′ adjacent to the UmCn in themRNA, and preference for the base 5′ adjacent to UmCn in themRNA being a C (and so a G at the 5′ end of the stem; Fig. 3A,Upper, and Fig. S5). Our experimental observations, taken to-gether with the previously identified structural and functional

properties of ternary elongation complexes, suggest a mecha-nistic model of the programmed hairpin-dependent generationof mRNA lacking one C from TmCn heteropolymeric motifs.

The Site and Directionality of Transcript Realignment. The com-ponents of the T5C5 motif are designated as follows:5′-dT1T2T3T4T5C1C2C3C4C5-3′ in the DNA nontemplate strandor 3′-dA5G5-5′ 3′-dA1A2A3A4A5G1G2G3G4G5-5′ in the DNAtemplate strand. Referring to the DNA template strand, re-alignment of the RNA-DNA hybrid leading to lack of a rC canoccur at positions G1, G2, G3, or G4 with a corresponding RNA3′ end having a C1, C2, C3, or C4 tract, respectively. Irrespective

Fig. 6. Individual steps in programmed transcriptional realignment at T5C5sequence. (Upper Right) Realignment site possibilities with nascent RNAshaving C1, C2, C3, or C4 at their 3′ ends. The sole position of the rU5:dG1

mispair in the possible newly realigned hybrids is indicated by an opensquare. (Upper Left) the cartoon shows individual functional blocks of theT5C5 motif: the RNA stem loop-forming sequence (green), 5′GC3′ base pair(brown) in the hairpin stem immediately adjacent to the T5 tract (blue), theC5 tract (red), and pyrimidine residue (T or C) in the template DNA strand(black) at the 3′ of the C5 tract. The realignment occurs at the C4 site of theC5 tract and at a 9-nt distance from the rG:rC base pair at the base of thestem (shown by a double-headed arrow). Sequence elements on the top arenecessary for realignment. (Lower) The putative structures and rearrange-ments of the RNA-DNA hybrid in RNAP paused at the C4 position of the T5C5motif. The partitions of the RNAP active center, the RNA 3′ end-binding site(i) and the NTP binding site (i + 1) are shown in magenta. Arrows indicatematches, alternative base pairs, and mismatches in the RNA-DNA hybridemerging at each step of the realignment. (Lower Right) Scheme depicts trans-location states of RNAP in each slippage intermediate (post- or pretranslocated)where n, n + 1, and n + 2 correspond to RNAP with the 3′ RNA end base pairswith G4, G5, and 3′Pu (G/A) template DNA position, respectively. Double-headedarrows indicate two branching points in the realignment pathway, which areaffected by CTP and UTP concentrations. High CTP/low UTP inhibits and highUTP/low CTP stimulates the realignment at C4 site. See Discussion for details.

E1990 | www.pnas.org/cgi/doi/10.1073/pnas.1418384112 Penno et al.

Dow

nloa

ded

by g

uest

on

May

24,

202

1

of location, the shift involves an rU5:dG1 mispair in the newlyrealigned RNA-DNA hybrid (the RNA bases specified by themotif have the same subscript number as in the template DNAmotif sequence). The distance from the rU5:dG1 mismatch to theRNAP catalytic center can extend from 1 to 4 rC:dG bp, de-pending on the slippage site, so a shift at G4 would involve asignificantly more stable hybrid in the RNAP catalytic centerthan a shift at G1 (Fig. 6, Upper Right).Efficient generation of transcripts lacking one C from the

TmCn motif (m + n = 10) variants with sequential combinationsfrom T1C9, T2C8. . . to T7C3 is evidence that the presence ofone rU:dG mismatch at any of the first eight hybrid RNA:DNAbase pairing positions of the realigned hybrid are stabilized bystructural features of the RNAP. However, realignment in-volving an rU:dG mismatch in the 9th or 10th hybrid positionsdoes not yield transcripts lacking one C, permitting the de-duction that these mismatch positions are not similarly stabilized[suggesting that relevant realignment occurs with T9C1 is thereality of realignment in at least the other direction (−1) where amismatch is not involved at the same RNA base position, andtranscripts are generated that contain an additional U (Fig. 4,insert 38; reaction acyC).Mismatch stabilization at a subset of positions is supported by

the increased slippage propensity exhibited by mutants in thefork loop domain of the β subunit, which interacts with the thirdand fourth RNA base from the RNA 3′ end (22) (SI Results andFig. S2F). The suggested stability lock is consistent with X-ray andbiochemical analysis of RNAP elongation complexes that showtight contacts of the enzyme with the RNA-DNA hybrid that aremostly limited to the first 2–3 3′-proximal bps of the hybrid (11,40). Consequently, complete inhibition of slippage generatingRNA lacking one C with T8C2 (5′-dT1T2T3T4T5T6T7T8C1C2-3′),but not with the T7C3 motif (5′-dT1T2T3T4T5T6T7C1C2C3-3′),indicates that RNAP loses its tolerance to rU:dG mispairing inthe hybrid when located one rC:dG base pair from the RNA 3′end. It further implies that realignment occurs at position C2 forT7C3, but not at its other possible shift site C1 (T8C2 has onepossible shift site, C1). In addition, the T5 position of the T5C5motif can also be excluded as a shift site because realignment ofthe 3′end of the RNA at this site generates rU5:dG1 in the activecenter of RNAP, which significantly slows down the next nu-cleoside monophosphate (NMP) incorporation (37, 41, 42).Earlier it was shown that the active center in RNAP has in-trinsically low tolerance to even a minor anomaly in the RNA-DNA hybrid (40).Altering the base at the C5 position in the T1T2T3T4T5C1

C2C3C4C5T motif abolishes slippage, indicating that a C at thisposition is critical for realignment. Further, slippage efficiency isstrongly dependent on UTP incorporation; U is specified 3′adjacent to the C5 tract. These results for the T5C5T motif,taken together with those from the effect of positioning of therU:dG mismatch in TmCn motifs in which m + n = 10, indicatesthat realignment must involve the two last (9th and 10th) motifpositions. In summary, the TEC that appears to be the preferredcandidate for realignment at the 3′-gA1AAAA5G1GGG4G5a-5′template sequence has the 3′ end of the RNA, C4, base pairedwith underlined template G4; this is designated TECC4.Posttranslocated TECC4 contains the 9-bp RNA-DNA hybrid

(43) with its active site vacated from the RNA 3′ end andavailable for binding of the next cognate NTP (CTP) (Fig. 6,Left, TECC4). In the regular elongation pathway, this complexincorporates CMP to generate the pretranslocated TECC5 (Fig. 6,Right, Productive pathway TECC5), which continues transcriptionbeyond the slippery motif to synthesize fully complementary mRNA(U5C5). We propose that the transcript realignment pathwaycompetes with CMP incorporation in TECC4. Slippage moves the3′ end of the transcript rU5C4 from position i (RNA 3′ endbinding site) to position i + 1 (NTP binding site; Fig. 6, TECC4

and TECC5slip). Such a shift would convert the posttranslocated

TECC4 to the pretranslocated TECC5slip (the italics and subscript

indicate the nonstandard origin of the complex) without trans-

location of RNAP along the DNA or phosphodiester bond for-mation (consequent to substrate incorporation). In TECC5

slip, thenascent transcript (5′-rU1U2U3U4U5C1C2C3C4-3′end) base pairswith 3′-dA1A2A3A4A5G1G2G3G4G5-5′ in the template, gener-ating a relatively stable rU5:dG1 alternative base pair in themiddle of the RNA-DNA hybrid. A shift of the transcript inthe opposite direction (which would lead to a C insertion in thetranscript) would generate an unstable rC1:dA5 mismatch. Thebetter fit of rU5:dG1 compared with rC1.dA5 establishes slippagedirectionality (Fig. 6, TECC5

slip dG1:rU5). Note that our exper-iments did not address the mechanistic details of the transcriptrealignment process. Some possibilities are in SI Discussion.These possibilities include transcript realignment by defect/bulgemigration, i.e., consecutive dissociation and repairing of singlebase pairs within the initial RNA-DNA hybrid (44, 45), and thepotential role in stem loop-dependent slippage of RNAP hyper-forward translocation (46), with consequent decreased hybridlength and therefore stability (Fig. S6, hyper-translocation model).

RNAP Translocation and Pausing in the Generation of RNA Lacking aSingle C. TECC5

slip may re-enter the regular elongation pathwayby undergoing forward translocation, incorporating the nextcognate UMP, and generating mRNA lacking one C (Fig. 6, TECescaped). Alternatively, TECC5

slip can switch back to the re-gular TECC4 by reversal of the RNA shift (Fig. 6, Right, Re-align reversal TECC4). Therefore, fast translocation of TECC5

slipand fast incorporation of the next UMP would promote absenceof one C (Fig. 6, TECC5

slip and TEC escaped). The C5/T junction3′ adjacent to the T5C5 motif likely promotes rapid escapeof TECC5

slip from the slippage site, because 3′ Py/Py junctions(referring to the nontemplate strand) strongly promote forwardtranslocation of E. coli RNAP and yeast RNAPII, whereas Py/Pujunctions often induce transcription pausing at Py residues (37).Consistent with this notion, absence of a single C is inhibited whenpurine residues are specified 3′ adjacent to the TmCn motif.RNA:DNA mismatches at the −10 position of the RNA-DNAhybrid also promote RNAP translocation (37). Apparently, thedistal G–proximal C (Fig. 3, inserts 2, 5, and 51) preference in theend of the hairpin stem suggests that the rC:dA mismatch ismore efficient in promoting forward translocation of TECC5

slipthan the rG:dA mismatch.A slow translocation and delayed incorporation of the next

CMP by TECC4 are expected to favor the pathway leading to oneC absence compared with the regular elongation pathway. Tran-sient pausing at C4 could be supported by the highly biasedpurine/pyrimidine composition of its RNA-DNA hybrid. Pyrim-idine-rich (U-rich) nascent RNA was shown to stimulate pausingof E. coli RNAP (40, 47). The RNA hairpin may additionallycontribute to the pausing either by melting the RNA-DNA hy-brid in TECC4, by interacting with the RNAP, or by inducing ahyper-translocation of RNAP as proposed previously (33, 46, 48)(SI Discussion and Fig. S6). Our model above and below explainshow the prevalence of C absence depends on CTP and UTPconcentrations in vitro. At low [CTP], RNAP dwells at the C4position, providing additional time for the RNA shift (TECC4 →TECC5

slip) to occur. Rapid incorporation of UMP by TECC5slip at

high [UTP] competes with conversion of this complex back toTECC4. Thus, the single C absence is most prevalent in vitro athigh [UTP] and low [CTP] (Fig. 6, Right).

The Stem-Loop Structure in Transcriptional Slippage: Parallels withIntrinsic Termination. The realignment precursor TECC4 containsthe RNA hairpin with a 6-bp stem 5′ adjacent to the 9-bp hybrid(rU5C4:dA5G4). The RNA hairpin promotes melting of theupstream part of the RNA-DNA hybrid during transcriptiontermination (33, 37) [the Roseiflexus slippage stem loop is similarto the T3 intrinsic terminator of themetY-nusA-infB operon (49);Fig. 3I]. Interestingly, with phage T7 single-subunit RNAP, anRNA stem loop structure (TΦ) stimulates intrinsic terminationon the upstream RNA-DNA hybrid containing poly(A) or polyUtracts. In the absence of a stem loop, T7 RNAP slips on the same

Penno et al. PNAS | Published online April 6, 2015 | E1991

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

24,

202

1

motifs leading to nt addition(s), but this slippage is stronglyinhibited by stem loop formation (50). In vitro, a small fractionof the TECC4 also terminates transcription. Similar to slippageyielding absence of a single C, termination at C4 occurs in a stemloop-dependent manner. Therefore, within this sequence con-text, the hairpin-mediated melting of the upstream part of theRNA-DNA hybrid may have two alternative outcomes: termi-nation of the elongation complex or transcript realignment. Theproposed similarity of these mechanisms is further supported byin vitro inhibition of slippage on TmCn by the antiterminationprotein N (Fig. 5). N is thought to stabilize the RNA-DNAhybrid in RNAP (23, 37). In summary, programmed transcriptrealignment and intrinsic transcription termination may be con-sidered as two branches of the same pathway, both including de-stabilization of the RNA-DNA hybrid within RNAP. The relativeefficiencies of these two processes should depend on the RNA-DNA hybrid stability in the TmCn motif.

Mechanistic Summary. Slippage on the TmCn motif involves a pro-grammed unidirectional shift of the RNA strand of the RNA-DNAhybrid in RNAP, which occurs uniquely at the C4 position of theT5C5 motif. The productive unidirectional forward 1-bp shift at theTmCn motifs is secured by the stability of the alternative new-realigned hybrid(s) occurring at different sites in the TmCn sequence.Inviability of more than one rU:dG and even one rC:dA mismatchprevents a two-base deletion and any insertions, respectively. Mis-matches cause rapid reversal of the single nt backward shift. TheC5T (Py/Py) antipausing junction sequence, as opposed to the pause-inducing Py/Pu sequence, at the downstream end of the T5C5 motifpromotes rapid escape of the realigned complex to elongation be-fore reversal of the slippage. This mechanism implies that transienttranscription realignment occurs on a broad variety of hetero-polymeric sequences. However, rapid reversal of the realignmentprevents deletions and insertions in the mRNA unless additionalsequence elements that inhibit the reversal are present.

Perspective. The term programmed transcriptional realignment(PTR) for structure-mediated slippage seems merited. It evokes aparallel with programmed ribosomal frameshifting, especially withthose cases where a recoding signal involves interaction of thenascent peptide within components of the peptide exit channel ofthe ribosome. A different type of parallel is where there is sug-gestive evidence for mRNA structure formation within a translatingribosome (51). The present work opens a new perspective for po-tential slippage-mediated regulation and protein diversity. Futurechallenges include testing initial bioinformatics-derived candidatesand extending the searches to include eukaryotic organisms.

Materials and MethodsBacteria Strains and Plasmids. E. coli strains and plasmids and oligonucleo-tides (IDT DNA) are in Tables S1–S3. The sequence of the inserts is also in-dicated in Fig. 3.

RNA Purification, Reverse Transcription, and PCR Reactions. Strains grown inLuria Bertani (LB) media to midlog phase were induced, when required, with1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for 20 min. RNA wereisolated using TRIzol (Invitrogen), treated with DNaseI turbo (Ambion), andpurified with a base silica column (Anachem). Reverse transcription using1 μg total RNA and SuperScriptIII reverse transcriptase (Invitrogen) was at52 °C. The sequence of the R#669 primer used on vector pJ307 gst-mbp wasGGTGACTTTAATTCCGGTATC (this anneals 102 nt downstream the BamHI 3′cloning site) and of R#30 on vector pJ123 GST-lacZ, was ACGACGTTG-TAAAACGACGG (anneals 22nt downstream the BglII 3′ cloning site). PCRreactions, 50 μL, with 2 μL reverse transcription reaction product or plasmid

DNA, were performed with primers R#30 and F#450 (5′-CCCAATGTGCCTG-GATGCG-3′) and with R#669 and F#668 (5′-TAAGTACTTGAAATCCAGCAAG-3′).PCR reactions contained Taq DNA polymerase 0.8 units (Biolabs), 1× thermobuffer (Biolabs), 200 μM each dNTP (Biolabs), 500 nM each primer, and 2 μLcDNA reaction or 0.1–0.5 ng plasmid. As a control for DNA contamination, re-verse transcription reactions were performed without reverse transcriptase andused as template for PCR reactions. Reaction amplification involved 25 cycles,with 30-s denaturation at 94 °C, 30-s annealing at 52 °C, and 30-s elongation at72 °C. See Fig. 2 for the strategy used for transcriptional slippage analysis.

LPE Reactions. RT-PCR and PCR products were agarose gel purified with a PCRclean up kit (Anachem). Primers with IRD700 fluorescent labeling wereIRD700-R#226 (pJ123) TAGGGCCCTTCGAAGATCTA and IRD700-R#821(pJ307) CAGTTGGGAATTCTTGGATCTA. Reactions, 12.5 μL, contained 0.05pmol DNA template, 0.15 pmol IRD700 labeled oligo, 1 μM final of threedNTP (Biolabs) with the missing one replaced by 50 μM of the correspondingacyNTP (Biolabs), and 0.05 U/μL of Vent exo− DNA polymerase (Biolabs).Reaction were linearly amplified for 60 cycles, with 30-s denaturation at94 °C, 30-s annealing at 55 °C, and 30-s elongation at 72 °C. Products wereanalyzed on a 15% urea sequencing gel. Signal detections and image cap-tures were performed by means of a Li-Cor sequencing instrument (Li-CorBiosciences). Band signal intensity was quantified using ImageQuant. Fig. S7depicts optimization of the LPE reactions conditions.

Protein Analysis. Pulse chase experiments were performed as in ref. 52 in Mops(morpholinepropanesulfonic acid)-glucose containing 100 μg/mL ampicillin andall amino acids (150 μg/mL each) except methionine and tyrosine, pulsed with[35S]methionine for 2 min, and chased with unlabeled Met for 2 min; the ex-ception, where indicated, is for induction with IPTG, where it was for 20 min.Proteins were separated on 10% SDS/PAGE. Dried gels were exposed to aMolecular Dynamics PhosphorImager screen, and band intensity was quanti-fied by ImageQuant. The frameshifting efficiency was estimated as in ref. 53.

In Vitro Experiments. RNA polymerase from E. coli and RNAP II from S. cer-evisiae were purified as described previously (41, 54). NTPs were purified asdescribed in ref. 41. Bacteriophage lambda N protein was purified as describedin ref. 23. Details of transcription procedures are provided in SI Materials andMethods. Briefly, to prepare the initial TEC (Fig. 5A), the TEC was first as-sembled from synthetic RNA and DNA oligonucleotides (35) and then ligatedto a DNA test insert. The TEC was walked by addition of NTP subsets to form a63 nt transcript stalled 8 nt upstream of the slippery site (3′-A5G5-5′) on theDNA template. The 3′ CMP of nascent RNA in this TEC is hybridized to dGwithin the 3′-CGTT-5′ sequence, which specified the loop of the RoseiflexusRNA structure. The 63 nt nascent RNA was digested with RNase T1, producinga 14 nt nascent RNA (5′-ACGGACGCGGGCGC-3′). The RNA was labeled by in-cubation with α-[32P] ATP (3,000 mCi/mol), producing a 16 nt RNA. The initialTEC carrying 22 nt RNA was obtained after an additional walking step.

The initial TEC was incubated with 50 μMUTP and GTP and 5, 50, or 500 μMCTP for 4 min at ambient temperature. For the experiments with the T5C5tagtemplate, GTP was substituted with ATP. When indicated, the initial TEC waspreincubated with 100-fold molar excess of N protein for 10 min. Reactionswere stopped by adding the denaturing gel loading buffer containing EDTAand urea, and the reactions products were separated in a 10% sequencing gel.Gels were exposed to a Molecular Dynamics PhosphorImager screen, and bandintensity was quantified by ImageQuant. Slippage efficiency was estimated asthe fraction of slippage product of the sum of the slippage and standard products.

ACKNOWLEDGMENTS. We thank D. J. Jin, D. Friedman, and M. Cashel forbacterial strains; G. Loughran, L. Renault, and S. Rinke for support; H. Feldmannand M. Mehedi for sending the coding sequence information of Ebola viruscassettes used in their work; and O. Fayet for stimulating us to investigate ISelements and a rewarding collaboration. This work was supported by an IrishResearch Council fellowship (to C.P.), Science Foundation Ireland Grants 13/IA/1853, 07/CE/B1365, and 12/RC/2273 (to J.F.A. and D.v.S.), Wellcome Trust Grant094423 (to P.V.B.), an Irish Health Research Board postdoctoral fellowshipGrant PDTM/2011/09 (to M.O.M.), a National Institutes of Health intramuralaward (to M.K.), and a University College Cork seed grant (to J.F.A.).

1. Larsen B, Wills NM, Nelson C, Atkins JF, Gesteland RF (2000) Nonlinearity in

genetic decoding: Homologous DNA replicase genes use alternatives of tran-

scriptional slippage or translational frameshifting. Proc Natl Acad Sci USA 97(4):

1683–1688.2. Antonov I, Coakley A, Atkins JF, Baranov PV, Borodovsky M (2013) Identification of

the nature of reading frame transitions observed in prokaryotic genomes. Nucleic

Acids Res 41(13):6514–6530.

3. Penno C, Sansonetti P, Parsot C (2005) Frameshifting by transcriptional slippage is

involved in production of MxiE, the transcription activator regulated by the ac-

tivity of the type III secretion apparatus in Shigella flexneri. Mol Microbiol 56(1):

204–214.4. Penno C, et al. (2006) Transcriptional slippage controls production of type III

secretion apparatus components in Shigella flexneri. Mol Microbiol 62(5):

1460–1468.

E1992 | www.pnas.org/cgi/doi/10.1073/pnas.1418384112 Penno et al.

Dow

nloa

ded

by g

uest

on

May

24,

202

1

5. Gueguen E, Wills NM, Atkins JF, Cascales E (2014) Transcriptional frameshifting res-cues Citrobacter rodentium type VI secretion by the production of two length variantsfrom the prematurely interrupted tssM gene. PLoS Genet 10(12):e1004869.

6. Baranov PV, Fayet O, Hendrix RW, Atkins JF (2006) Recoding in bacteriophages andbacterial IS elements. Trends Genet 22(3):174–181.

7. Baranov PV, Hammer AW, Zhou J, Gesteland RF, Atkins JF (2005) Transcriptionalslippage in bacteria: Distribution in sequenced genomes and utilization in IS elementgene expression. Genome Biol 6(3):R25.

8. Tamas I, et al. (2008) Endosymbiont gene functions impaired and rescued by polymeraseinfidelity at poly(A) tracts. Proc Natl Acad Sci USA 105(39):14934–14939.

9. Wernegreen JJ, Kauppinen SN, Degnan PH (2010) Slip into something more func-tional: Selection maintains ancient frameshifts in homopolymeric sequences. Mol BiolEvol 27(4):833–839.

10. Bar-Nahum G, et al. (2005) A ratchet mechanism of transcription elongation and itscontrol. Cell 120(2):183–193.

11. Vassylyev DG, Vassylyeva MN, Perederina A, Tahirov TH, Artsimovitch I (2007)Structural basis for transcription elongation by bacterial RNA polymerase. Nature448(7150):157–162.

12. Wagner LA, Weiss RB, Driscoll R, Dunn DS, Gesteland RF (1990) Transcriptional slip-page occurs during elongation at runs of adenine or thymine in Escherichia coli.Nucleic Acids Res 18(12):3529–3535.

13. Uptain SM, Kane CM, Chamberlin MJ (1997) Basic mechanisms of transcript elonga-tion and its regulation. Annu Rev Biochem 66:117–172.

14. Chamberlin M, Berg P (1962) Deoxyribo ucleic acid-directed synthesis of ribonucleicacid by an enzyme from Escherichia coli. Proc Natl Acad Sci USA 48(1):81–94.

15. Thomas SM, Lamb RA, Paterson RG (1988) Two mRNAs that differ by two non-templated nucleotides encode the amino coterminal proteins P and V of the para-myxovirus SV5. Cell 54(6):891–902.

16. Jacques JP, Kolakofsky D (1991) Pseudo-templated transcription in prokaryotic andeukaryotic organisms. Genes Dev 5(5):707–713.

17. Lamb RA, Parks GD (2007) Paramyxoviridae: The viruses and their replication. FieldsVirology, eds Knipe DM, Howley PM (Lippincott Williams & Wilkins, Philadelphia),5th Ed, pp 1449–1496.

18. Anikin M, Molodtsov V, Temiakov D, McAllister WT (2010) Transcript Slippage andRecoding. Recoding: Expansion of Decoding Rules Enriches Gene Expression, edsAtkins JF, Gesteland RF (Springer, New York), pp 409–432.

19. Kulkarni S, et al. (2009) Nipah virus edits its P gene at high frequency to express the Vand W proteins. J Virol 83(8):3982–3987.

20. Iseni F, et al. (2002) Chemical modification of nucleotide bases and mRNA editingdepend on hexamer or nucleoprotein phase in Sendai virus nucleocapsids. RNA 8(8):1056–1067.

21. Kolakofsky D, Roux L, Garcin D, Ruigrok RWH (2005) Paramyxovirus mRNA editing,the “rule of six” and error catastrophe: A hypothesis. J Gen Virol 86(Pt 7):1869–1877.

22. Zhou YN, et al. (2013) Isolation and characterization of RNA polymerase rpoB mu-tations that alter transcription slippage during elongation in Escherichia coli. J BiolChem 288(4):2700–2710.

23. Parks AR, et al. (2014) Bacteriophage λ N protein inhibits transcription slippage byEscherichia coli RNA polymerase. Nucleic Acids Res 42(9):5823–5829.

24. Hausmann S, Garcin D, Delenda C, Kolakofsky D (1999) The versatility of para-myxovirus RNA polymerase stuttering. J Virol 73(7):5568–5576.

25. Jacques JP, Hausmann S, Kolakofsky D (1994) Paramyxovirus mRNA editing leads to Gdeletions as well as insertions. EMBO J 13(22):5496–5503.

26. Sharma V, et al. (2011) A pilot study of bacterial genes with disrupted ORFs reveals asurprising profusion of protein sequence recoding mediated by ribosomal frame-shifting and transcriptional realignment. Mol Biol Evol 28(11):3195–3211.

27. Mejlhede N, et al. (2004) -1 frameshifting at a CGA AAG hexanucleotide site is re-quired for transposition of insertion sequence IS1222. J Bacteriol 186(10):3274–3277.

28. Rettberg CC, Prère MF, Gesteland RF, Atkins JF, Fayet O (1999) A three-way junctionand constituent stem-loops as the stimulator for programmed -1 frameshifting inbacterial insertion sequence IS911. J Mol Biol 286(5):1365–1378.

29. Zwieb C (1992) Conformity of RNAs that interact with tetranucleotide loop bindingproteins. Nucleic Acids Res 20(17):4397–4400.

30. Weiss RB, Dunn DM, Atkins JF, Gesteland RF (1987) Slippery runs, shifty stops, back-ward steps, and forward hops: -2, -1, +1, +2, +5, and +6 ribosomal frameshifting. ColdSpring Harb Symp Quant Biol 52:687–693.

31. Kashkina E, et al. (2007) Multisubunit RNA polymerases melt only a single DNA basepair downstream of the active site. J Biol Chem 282(30):21578–21582.

32. Kireeva ML, Domecq C, Coulombe B, Burton ZF, Kashlev M (2011) Interaction of RNApolymerase II fork loop 2 with downstream non-template DNA regulates transcrip-tion elongation. J Biol Chem 286(35):30898–30910.

33. Komissarova N, Becker J, Solter S, Kireeva M, Kashlev M (2002) Shortening of RNA:DNA hybrid in the elongation complex of RNA polymerase is a prerequisite fortranscription termination. Mol Cell 10(5):1151–1162.

34. Toulokhonov I, Artsimovitch I, Landick R (2001) Allosteric control of RNA polymeraseby a site that contacts nascent RNA hairpins. Science 292(5517):730–733.

35. Komissarova N, Kireeva ML, Becker J, Sidorenkov I, Kashlev M (2003) Engineering ofelongation complexes of bacterial and yeast RNA polymerases.Methods Enzymol 371:233–251.

36. Strathern J, et al. (2013) The fidelity of transcription: RPB1 (RPO21) mutations thatincrease transcriptional slippage in S. cerevisiae. J Biol Chem 288(4):2689–2699.

37. Imashimizu M, Shimamoto N, Oshima T, Kashlev M (2014) Transcription elongation:Heterogeneous tracking of RNA polymerase and its biological implications. Tran-scription 5:5.

38. Artsimovitch I, Landick R (2000) Pausing by bacterial RNA polymerase is mediated bymechanistically distinct classes of signals. Proc Natl Acad Sci USA 97(13):7090–7095.

39. Mehedi M, et al. (2013) Ebola virus RNA editing depends on the primary editing sitesequence and an upstream secondary structure. PLoS Pathog 9(10):e1003677.

40. Bochkareva A, Yuzenkova Y, Tadigotla VR, Zenkin N (2012) Factor-independenttranscription pausing caused by recognition of the RNA-DNA hybrid sequence. EMBOJ 31(3):630–639.

41. Kireeva ML, et al. (2008) Transient reversal of RNA polymerase II active site closingcontrols fidelity of transcription elongation. Mol Cell 30(5):557–566.

42. Imashimizu M, Oshima T, Lubkowska L, Kashlev M (2013) Direct assessment of tran-scription fidelity by high-resolution RNA sequencing. Nucleic Acids Res 41(19):9090–9104.

43. Vassylyev DG, et al. (2002) Crystal structure of a bacterial RNA polymerase holoen-zyme at 2.6 A resolution. Nature 417(6890):712–719.

44. Banavali NK (2013) Partial base flipping is sufficient for strand slippage near DNAduplex termini. J Am Chem Soc 135(22):8274–8282.

45. Neher RA, Gerland U (2004) Dynamics of force-induced DNA slippage. Phys Rev Lett93(19):198102.

46. Nedialkov YA, Nudler E, Burton ZF (2012) RNA polymerase stalls in a post-translocatedregister and can hyper-translocate. Transcription 3(5):260–269.

47. Gusarov I, Nudler E (1999) The mechanism of intrinsic transcription termination. MolCell 3(4):495–504.

48. Hein PP, et al. (2014) RNA polymerase pausing and nascent-RNA structure formationare linked through clamp-domain movement. Nat Struct Mol Biol 21(9):794–802.

49. Bylund GO, Nord S, Lövgren JM, Wikström PM (2011) Alterations in the β flap and β’dock domains of the RNA polymerase abolish NusA-mediated feedback regulation ofthe metY-nusA-infB operon. J Bacteriol 193(16):4113–4122.

50. Molodtsov V, Anikin M, McAllister WT (2014) The presence of an RNA:DNA hybridthat is prone to slippage promotes termination by T7 RNA polymerase. J Mol Biol426(18):3095–3107.

51. Atkins JF, Gestland RF, eds (2010) Recoding: Expansion of Decoding Rules EnrichesGene Expression (Springer, New York), pp 1–466.

52. Gurvich OL, et al. (2003) Sequences that direct significant levels of frameshifting arefrequent in coding regions of Escherichia coli. EMBO J 22(21):5941–5950.

53. Grentzmann G, Ingram JA, Kelly PJ, Gesteland RF, Atkins JF (1998) A dual-luciferasereporter system for studying recoding signals. RNA 4(4):479–486.

54. Kireeva ML, Kashlev M (2009) Mechanism of sequence-specific pausing of bacterialRNA polymerase. Proc Natl Acad Sci USA 106(22):8900–8905.

Penno et al. PNAS | Published online April 6, 2015 | E1993

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

24,

202

1